Exploration Logs

File §

• ¹ propose an ABM network that is compatible with the use of MARL . The framework encodes the following
  • Partial Observability.
  • A network model for inter-agent relationships. Connectivity can either be static or stochastic.
  • Agent Utility Functions encapsulated as types.
  • Heterogeneous Agent Preferences
  • Support for complex turn orders (i.e., turns based on types)

• Games can become competitive for their own good — driving the skill floor and skill ceiling up without an influx of players to make playing sustainable for casuals.

Queue §

• ² provides a theoretical and empirical analysis of the use of Centralized Critics in CTDE.

• ³ introduces a new mutual information framework for MARL. This leads to the development of an algorithm called Variational Maximum Mutual Information, Multi-Agent Actor Critic which allows agents to coordinate simultaneous actions without latency.

• RAMBO Attack - an attack involving writing to RAM and picking up on the EM signals emitted.

• Branching Reinforcement Learning by Du, and Chen (Jun 15, 2022)

• Vinyals et al. (2019) Grandmaster level in StarCraft II using multi-agent reinforcement learning

  • Linked in Self Play

• Wu et al. (2017) Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation

  • Linked in Trust Region Policies

• Ecoclimates — Climate-Response Modeling of Vegetation by Palubicki et al. (2022)

• Ma et al. (2024) Foundation Methods for Music — A survey

Footnotes §

1. Ardon et al. (2023) An RL driven multi-agent framework to model complex systems ↩

2. Lyu et al. (2023) On Centralized Critics in Multi-Agent Reinforcement Learning ↩

3. Kim, Jung, Cho, Sung (2020) A Maximum Mutual Information Framework for Multi-Agent Reinforcement Learning ↩